Datensatz

Zusammenfassung

This thesis presents the development and implementation of boundary potentials for hybrid quantum mechanical/molecular mechanical (QM/MM) methods. The dual-layer QM/MM method is extended to a three-layer method with the objective of providing an accurate and efficient description of long range electrostatic interactions.
First, a generalized solvent boundary potential (GSBP) originally developed for classical force field simulations is adapted for hybrid QM/MM methods with semiempirical QM Hamiltonians. The GSBP is tested on a model system (threonine in water) and is found to yield accurate results. The efficiency of the GSBP is studied and the breakeven point with standard QM/MM calculations is located at system sizes of around 12,500 atoms. The GSBP reduces the computational costs by 70 % for systems with about 25,000 atoms which are common in QM/MM studies. Since application of the GSBP is connected with a significant overhead, three algorithmic improvements are introduced that reduce the computation time of the overhead by 60 % with only minimal loss of accuracy.
Thereafter, a novel solvated macromolecule boundary potential (SMBP) is introduced which, in contrast to the GSBP, targets geometry optimizations and can be applied with density functional theory or ab initio methods for the QM region. The SMBP is conceptually similar to the GSBP: The outer macromolecule region is represented by a boundary potential obtained from solution of the Poisson-Boltzmann equation; the outer solvent molecules are modeled as a dielectric continuum. A modular implementation that allows application with any QM/MM Hamiltonian is achieved by combining a self consistent reaction field procedure with a point charge-based representation of the boundary potential in the QM calculations. The SMBP is tested on a model system (glycine in water) and three enzymatic systems (p-hydroxybenzoate hydroxylase, cytochrome P450cam, and chorismate mutase). In the case of solvated glycine, application of the SMBP turns out to be problematic since QM/MM and QM/MM/SMBP optimizations lead to different local minima with different energetics despite their structural similarity. In the enzymatic systems, the SMBP reproduces the electrostatic potential with high accuracy and computed potential energy differences rarely deviate by more than 0.3 kcal/mol from the full QM/MM results. Molecular and electronic structures resulting from QM/MM/SMBP geometry optimizations can be used as input for free energy computations following the QM/MM-free energy perturbation scheme. The conceptual similarity of GSBP and SMBP permits application of the GSBP during configurational sampling thereby reducing the computational costs of this step by up to 90%.
Long range electrostatic interactions in enzymes can have two sources: the outer macromolecule and the surrounding solvent. The effect of both contributions on enzymatic reactions is studied by means of SMBP and GSBP. It is found that both contributions influence reaction energetics considerably only if there is significant charge transfer during the reaction. In such cases an accurate description of both contributions is necessary. GSBP and SMBP offer such accuracy at reduced computational costs.